Fetal cortical plate segmentation

Fetal cortical plate segmentation with multiple predictions using fully convolutional networks

• Free software: GNU General Public License v3

Abstract

With the development of fetal brain imaging technology, quantitative methods for evaluating fetal cortical development have been suggested. For reliable quantitative analysis of cortical volume and sulcal folding, accurate and automated segmentation of the cortical plate is first necessary. Accordingly, we propose a fully convolutional neural network for automatic segmentation of the cortical plate. We adopted a hybrid loss function to improve the segmentation accuracy, and multi-view (axial, coronal, sagittal) aggregation with test-time augmentation method to reduce errors caused by a single prediction and utilize 3D information. For evaluation of our method, 10-fold cross-validation of 52 fetuses brain MRI (22.9-31.4 weeks of gestation) was used. The proposed method obtained the Dice index 0.907 ± 0.027, 0.906 ± 0.031 and the mean surface distance error 0.182 ± 0.058mm, 0.185 ± 0.069mm for the left and right cortical plate respectively, and 0.980 ± 0.008, 0.979 ± 0.011 and 0.265 ± 0.090mm, 0.270 ± 0.110mm for inner volume of cortical plate. The results demonstrated that our method obtained higher accuracy in segmentation compared to other prior methods, and the proposed hybrid loss function and aggregation of multi-view with multiple predictions were effective in improving the accuracy.

Preprocessing

• High resolution reconstraction (isotropic 0.75mm voxel) using Kuklisova-Murgasova et al., 2012
• Linear registration to 31 weeks GA brain template using flirt in FSL.

Arguments

python3 fetal_cp_seg.py
        -h, --help            show this help message and exit
        -input INP, --input_MR INP [required]
                input MR file name ('.nii or .nii.gz') or folder name
        -output OUT, --output_loc OUT [required]
                Output path
        -axi AXI, --axi_weight AXI
                Axial weight file
        -cor COR, --cor_weight COR
                Coronal weight file
        -sag SAG, --sag_weight SAG
                Sagittal weight file

Run

This plugin can be run in two modes: natively as a python package or as a containerized docker image.

Using Python3

To run using native env or virtual env, simply do a

git clone https://github.com/jwhong1125/fetal_CP_segmentation.git
pip3 install -r requirements_dev.txt

and run with

python3 fetal_cp_seg.py -input {.nii or .nii.gz or folder} -output {output_path}
                        [-axi {pretraied axial view network weights}
                         -cor {pretraied coronal view network weights}
                         -sag {pretraied sagittal view network weights}]

Using docker run

To run using docker, be sure to assign an "input" directory to /incoming and an output directory to /outgoing. Make sure that the $(pwd)/out directory is world writable!

mkdir in out && chmod 777 out
docker run --rm -v $(pwd)/in:/incoming -v $(pwd)/out:/outgoing      \
        jwhong1125/fetal_cp_seg fetal_cp_seg.py                     \
        -input /incoming                                            \
        -ouput /outgoing

Binary package

For your convenience, we provide pre-compiled a binary file on ubuntu 16.04 or later. https://drive.google.com/file/d/1sYtdMqJoBjYpfqDjArxpnIYrvlqDNaYN/view?usp=sharing

Credits

This package was created with Cookiecutter and the audreyr/cookiecutter-pypackage project template.